diff options
author | Bjørn Erik Pedersen <[email protected]> | 2017-10-04 22:12:51 +0200 |
---|---|---|
committer | Bjørn Erik Pedersen <[email protected]> | 2017-10-07 18:00:07 +0200 |
commit | 5800a20a258378440e203a6c4a4343f5077755df (patch) | |
tree | dca1bd5030dccd56b4e9ad2dc1e037955aa4e20d /metrics | |
parent | e2e8bcbec34702a27047b91b6b007a15f1fc0797 (diff) | |
download | hugo-5800a20a258378440e203a6c4a4343f5077755df.tar.gz hugo-5800a20a258378440e203a6c4a4343f5077755df.zip |
metrics: Detect partialCached candidates
This commit adds a "cache potential" column when running `hugo --templateMetrics --templateMetricsHints`.
This is only calculated when `--templateMetricsHints` is set, as these calculations has an negative effect on the other timings.
This gives a value for partials only, and is a number between 0-100 that indicates if `partial` can be replaced with `partialCached`.
100 means that all execution of the same partial resulted in the same output.
You should do some manual research before going "all cache".
Diffstat (limited to 'metrics')
-rw-r--r-- | metrics/metrics.go | 145 | ||||
-rw-r--r-- | metrics/metrics_test.go | 50 |
2 files changed, 179 insertions, 16 deletions
diff --git a/metrics/metrics.go b/metrics/metrics.go index 9f8a158ad..c83610a92 100644 --- a/metrics/metrics.go +++ b/metrics/metrics.go @@ -17,7 +17,10 @@ package metrics import ( "fmt" "io" + "math" "sort" + "strconv" + "strings" "sync" "time" ) @@ -31,21 +34,48 @@ type Provider interface { // WriteMetrics will write a summary of the metrics to w. WriteMetrics(w io.Writer) + // TrackValue tracks the value for diff calculations etc. + TrackValue(key, value string) + // Reset clears the metric store. Reset() } +type diff struct { + baseline string + count int + simSum int +} + +func (d *diff) add(v string) *diff { + if d.baseline == "" { + d.baseline = v + d.count = 1 + d.simSum = 100 // If we get only one it is very cache friendly. + return d + } + + d.simSum += howSimilar(v, d.baseline) + d.count++ + + return d +} + // Store provides storage for a set of metrics. type Store struct { - metrics map[string][]time.Duration - mu *sync.Mutex + calculateHints bool + metrics map[string][]time.Duration + mu sync.Mutex + diffs map[string]*diff + diffmu sync.Mutex } // NewProvider returns a new instance of a metric store. -func NewProvider() Provider { +func NewProvider(calculateHints bool) Provider { return &Store{ - metrics: make(map[string][]time.Duration), - mu: &sync.Mutex{}, + calculateHints: calculateHints, + metrics: make(map[string][]time.Duration), + diffs: make(map[string]*diff), } } @@ -54,6 +84,32 @@ func (s *Store) Reset() { s.mu.Lock() s.metrics = make(map[string][]time.Duration) s.mu.Unlock() + s.diffmu.Lock() + s.diffs = make(map[string]*diff) + s.diffmu.Unlock() +} + +// TrackValue tracks the value for diff calculations etc. +func (s *Store) TrackValue(key, value string) { + if !s.calculateHints { + return + } + + s.diffmu.Lock() + var ( + d *diff + found bool + ) + + d, found = s.diffs[key] + + if !found { + d = &diff{} + s.diffs[key] = d + } + + d.add(value) + s.diffmu.Unlock() } // MeasureSince adds a measurement for key to the metric store. @@ -74,6 +130,12 @@ func (s *Store) WriteMetrics(w io.Writer) { var sum time.Duration var max time.Duration + diff, found := s.diffs[k] + cacheFactor := 0 + if found { + cacheFactor = int(math.Floor(float64(diff.simSum) / float64(diff.count))) + } + for _, d := range v { sum += d if d > max { @@ -83,31 +145,42 @@ func (s *Store) WriteMetrics(w io.Writer) { avg := time.Duration(int(sum) / len(v)) - results[i] = result{key: k, count: len(v), max: max, sum: sum, avg: avg} + results[i] = result{key: k, count: len(v), max: max, sum: sum, avg: avg, cacheFactor: cacheFactor} i++ } s.mu.Unlock() - // sort and print results - fmt.Fprintf(w, " %13s %12s %12s %5s %s\n", "cumulative", "average", "maximum", "", "") - fmt.Fprintf(w, " %13s %12s %12s %5s %s\n", "duration", "duration", "duration", "count", "template") - fmt.Fprintf(w, " %13s %12s %12s %5s %s\n", "----------", "--------", "--------", "-----", "--------") + if s.calculateHints { + fmt.Fprintf(w, " %9s %13s %12s %12s %5s %s\n", "cache", "cumulative", "average", "maximum", "", "") + fmt.Fprintf(w, " %9s %13s %12s %12s %5s %s\n", "potential", "duration", "duration", "duration", "count", "template") + fmt.Fprintf(w, " %9s %13s %12s %12s %5s %s\n", "-----", "----------", "--------", "--------", "-----", "--------") + } else { + fmt.Fprintf(w, " %13s %12s %12s %5s %s\n", "cumulative", "average", "maximum", "", "") + fmt.Fprintf(w, " %13s %12s %12s %5s %s\n", "duration", "duration", "duration", "count", "template") + fmt.Fprintf(w, " %13s %12s %12s %5s %s\n", "----------", "--------", "--------", "-----", "--------") + + } sort.Sort(bySum(results)) for _, v := range results { - fmt.Fprintf(w, " %13s %12s %12s %5d %s\n", v.sum, v.avg, v.max, v.count, v.key) + if s.calculateHints { + fmt.Fprintf(w, " %9d %13s %12s %12s %5d %s\n", v.cacheFactor, v.sum, v.avg, v.max, v.count, v.key) + } else { + fmt.Fprintf(w, " %13s %12s %12s %5d %s\n", v.sum, v.avg, v.max, v.count, v.key) + } } } // A result represents the calculated results for a given metric. type result struct { - key string - count int - sum time.Duration - max time.Duration - avg time.Duration + key string + count int + cacheFactor int + sum time.Duration + max time.Duration + avg time.Duration } type bySum []result @@ -115,3 +188,43 @@ type bySum []result func (b bySum) Len() int { return len(b) } func (b bySum) Swap(i, j int) { b[i], b[j] = b[j], b[i] } func (b bySum) Less(i, j int) bool { return b[i].sum > b[j].sum } + +// howSimilar is a naive diff implementation that returns +// a number between 0-100 indicating how similar a and b are. +// 100 is when all words in a also exists in b. +func howSimilar(a, b string) int { + + if a == b { + return 100 + } + + // Give some weight to the word positions. + const partitionSize = 4 + + af, bf := strings.Fields(a), strings.Fields(b) + if len(bf) > len(af) { + af, bf = bf, af + } + + m1 := make(map[string]bool) + for i, x := range bf { + partition := partition(i, partitionSize) + key := x + "/" + strconv.Itoa(partition) + m1[key] = true + } + + common := 0 + for i, x := range af { + partition := partition(i, partitionSize) + key := x + "/" + strconv.Itoa(partition) + if m1[key] { + common++ + } + } + + return int(math.Floor((float64(common) / float64(len(af)) * 100))) +} + +func partition(d, scale int) int { + return int(math.Floor((float64(d) / float64(scale)))) * scale +} diff --git a/metrics/metrics_test.go b/metrics/metrics_test.go new file mode 100644 index 000000000..5a5553ecd --- /dev/null +++ b/metrics/metrics_test.go @@ -0,0 +1,50 @@ +// Copyright 2017 The Hugo Authors. All rights reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package metrics + +import ( + "strings" + "testing" + + "github.com/stretchr/testify/require" +) + +func TestSimilarPercentage(t *testing.T) { + assert := require.New(t) + + sentence := "this is some words about nothing, Hugo!" + words := strings.Fields(sentence) + for i, j := 0, len(words)-1; i < j; i, j = i+1, j-1 { + words[i], words[j] = words[j], words[i] + } + sentenceReversed := strings.Join(words, " ") + + assert.Equal(100, howSimilar("Hugo Rules", "Hugo Rules")) + assert.Equal(50, howSimilar("Hugo Rules", "Hugo Rocks")) + assert.Equal(66, howSimilar("The Hugo Rules", "The Hugo Rocks")) + assert.Equal(66, howSimilar("The Hugo Rules", "The Hugo")) + assert.Equal(66, howSimilar("The Hugo", "The Hugo Rules")) + assert.Equal(0, howSimilar("Totally different", "Not Same")) + assert.Equal(14, howSimilar(sentence, sentenceReversed)) + +} + +func BenchmarkHowSimilar(b *testing.B) { + s1 := "Hugo is cool and " + strings.Repeat("fun ", 10) + "!" + s2 := "Hugo is cool and " + strings.Repeat("cool ", 10) + "!" + + for i := 0; i < b.N; i++ { + howSimilar(s1, s2) + } +} |